This may seem obvious or blasphemous, depending on how invested you are in hyping up AI. But it’s true: Generative AI is not a spell you can cast on your business to achieve instant and unlimited growth. It’s a tool; like any tool, it has its applications and limitations. The key to getting results from AI is to understand both.
In distribution, machine learning and large language models are ideal for automating traditionally time-consuming processes and quickly analyzing the massive amounts of data distributors generate daily. But rather than diving headfirst into the first AI-powered solution you see, you should start by gaining a realistic view of what the tech can do.
Since generative AI broke into the mainstream a couple of years ago, we’ve seen many examples of what can happen when someone implements AI before fully understanding it.
In New York, a lawyer with three decades of experience used ChatGPT for research and submitted a legal brief that included six fabricated cases. He later told the court that he was “unaware that [ChatGPT’s] content could be false.”
In California, a car dealership’s AI chatbot was tricked into selling a new car for $1. The sale wasn’t honored, of course, but viral mockery is certainly not the sort of publicity that a business hopes for.
Generative AI tools have also been observed to experience hallucinations, commit plagiarism, and even make prejudiced decisions about people.
These cases come down to expecting great output without giving enough consideration to the input. Artificial intelligence is only as intelligent as the data it’s trained on, and even the smartest AI tool will be limited by the quality of the data it’s made to analyze. To put it more simply, garbage in, garbage out.
McKinsey opens a 2023 report on AI by saying, “If your data isn’t ready for generative AI, your business isn’t ready for generative AI.” In other words, you can’t just install an AI tool right now and expect it to work seamlessly. First, you must ensure that your data is clean, well-organized, and easily accessible. If it’s not, your output will be garbage.
Distributors who want their AI-driven initiatives to succeed need to be on the lookout for:
- Bad inventory data. You could design the world’s greatest AI-powered demand forecasting tool, but if your inventory data isn’t accurate, all that predictive power will go to waste, and you’ll end up overstocking, tying up too much capital, or stocking out.
- Duplicate customer records. Whether from user error or system integrations, duplicate records are bound to throw off an AI tool’s customer behavior analysis and personalized marketing abilities.
- Inaccurate product information. If your product descriptions, specs, or prices are inaccurate or inconsistent, your shiny new AI sales recommendation platform will fail to impress customers.
For each of these, the key is having clean, accurate data housed within a solidly built IT infrastructure. Before implementing any AI-powered solution, you must implement rigorous data management practices, ensuring that your data has enough integrity for the solution’s output to be worthwhile.
Once your data is in good shape and your data architecture foundation can handle the heavy processing load of generative AI, you can focus on developing and training your AI tools. This involves curating an industry-, company-, and role-specific knowledge base from which those tools will “learn.” It also involves investing resources into prompt engineering, which means developing well-structured prompt formats that will help avoid source-reference divergences (which can cause hallucinations) and biased frameworks and ensure high-quality, on-task outputs.
Distributors looking to get value from generative AI will also need a strong and secure IT infrastructure. Suppose a mischievous troll can convince a chatbot to sell them a car for a dollar. In that case, a well-equipped hacker can manipulate an unprotected generative AI system for more nefarious ends. If you’re hoping to reap AI’s rewards, you must first invest in security measures that will protect sensitive business data from bad actors.
It may feel like I’m telling you to transform your distribution company into a tech company. But while distributors need to have a strong working knowledge of AI and other emerging technologies that offer the potential to spur growth, you don’t have to do it all on your own.
The best way to navigate treacherous yet promising terrain is to have an expert guide show you the way. Look for an AI software implementation partner that understands generative AI and can help you figure out how your business will leverage it. Then, you’ll have a much easier time avoiding those pitfalls and reaping the rewards.
Nick Pericle is the vice president of technology strategy and solutions at ProfitOptics.